Available online at www.sciencedirect.com

tters 262 (2007) 493–504www.elsevier.com/locate/epsl

Earth and Planetary Science Le

Seismicity variations associated with aseismic transients in Guerrero,Mexico, 1995–2006

Yajing Liu a,b,⁎, James R. Rice a,c, Kristine M. Larson d

a Department of Earth and Planetary Sciences, Harvard University, Cambridge, MA 02138, USAb Current affiliation: Department of Geosciences, Princeton University, Princeton, NJ 08544, USAc School of Engineering and Applied Sciences, Harvard University, Cambridge, MA 02138, USA

d Department of Aerospace Engineering Sciences, University of Colorado, Boulder, CO 80309, USA

Received 10 April 2007; received in revised form 27 July 2007; accepted 11 August 2007

Available onli

Editor: G.D. Price

ne 28 August 2007

Abstract

Primarily aseismic deformation transients in subduction zones, sometimes associated with tremors and low-frequency earthquakes,are a newly recognized mode of deformation. Stressing in the up-dip seismogenic zone is increased episodically due to down-diptransient slips, and each eventmaymake itmore prone to failure in a large thrust earthquake. It is important for seismic hazard assessmentto search and identify patterns of seismicity variation associated with transients. The Guerrero, Mexico, region is chosen for this studybecause of long-term continuous geodetic observations and abundant seismicity in the shallow subduction zone. We search the GCMTand NEIC catalogs for earthquakes with depths less than 100 km between 1995 and 2006within the area covering the region affected bymajor transients since 1996. A completeness magnitude ofMc=4.5 is determined for the NEIC catalog used in this study, based on themaximum likelihood method.

Three large transients in 1998, 2001–2002 and 2006 are all temporally correlated with high seismic rates in the studied area. Inparticular, transients are either preceded by a cluster of extensional earthquakes relatively far inland from the trench, or followed byshallow thrust earthquakes close to the trench. In some cases, such as the 2001–2002 transient, both types of activity are found borderingthe transient. The assembled evidence suggests that transients may serve as a mechanism of stress communication between distantseismicity clusters in shallow subduction zones.© 2007 Elsevier B.V. All rights reserved.

Keywords: aseismic transient; seismicity variation; Guerrero subduction zone

⁎ Corresponding author. Department of Geosciences, PrincetonUniversity, Princeton, NJ 08544, USA.

E-mail addresses: yjliu@princeton.edu (Y. Liu),rice@esag.harvard.edu (J.R. Rice), kristinem.larson@gmail.com(K.M. Larson).

0012-821X/$ - see front matter © 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.epsl.2007.08.018

1. Introduction

Aseismic transients with as yet no obvious relation tolarge earthquakes are a newly recognized mode of defor-mation along major fault zones. Short-period transients(several months to years) have recently been detected inshallow dipping subduction zones, such as Alaska–Aleutian (Ohta et al., 2006), Cascadia (Dragert et al.,2001; Rogers and Dragert, 2003; Szeliga et al., 2004;

494 Y. Liu et al. / Earth and Planetary Science Letters 262 (2007) 493–504

McCausland et al., 2005), Guerrero, Mexico (Lowryet al., 2001; Kostoglodov et al., 2003; Larson et al., 2004;Lowry, 2006), Hikurangi, New Zealand (Douglas et al.,2005; Wallace and Beavan, 2006), central and southwestJapan (Hirose et al., 1999; Ozawa et al., 2002; Hirose andObara, 2005), along the San Andreas Fault (Murray andSegall, 2005) and on a detachment beneath the southflank of Kilauea volcano (Segall et al., 2006). Natawid-jaja et al. (2004, 2007) also suggested the occurrence ofaseismic slip episodes, in 1962, 1968, 1975 and 1984, onthe Sunda Megathrust along the coast of Sumatra, basedon the annual banding of corals, and Meltzner et al.(2007) have noted coral evidence for an aseismic upliftevent in late 2003 in central Simeulue Island, althoughextreme climatic and oceanographic events can alsocontribute to the sea level fluctuations (A. Meltzner andK. Sieh, private communication). Seismic transients insome subduction zones are accompanied by deep non-volcanic tremors, which are difficult to locate due to lackof distinct P or S-wave arrivals. Tremors or low-frequency earthquakes, which are possibly componentsof tremor sequences, may be distributed over a broaddepth range (northern Cascadia, Kao et al., 2005), orclustered in a tabular zone along the thrust interface(Shikoku, SW Japan, Shelly et al., 2006) at depths wheretransient deformation is inferred to take place.

Transient deformation is estimated to be due to slowslips down-dip from the locked region. Stressing in theup-dip seismogenic zone is increased episodically, andeach transient can make it more prone to failure in a largethrust earthquake. Thus, it is very important for seismichazard assessment (Mazzotti and Adams, 2004) to searchand identify patterns of spatiotemporal seismicity varia-tion associated with transients. Such patterns are indeeddetected for some regions. For example, Segall et al.(2006) reported swarms of high frequency earthquakes,which accompanied, and are hypothesized to be triggeredby silent slip events on Kilauea volcano, Hawaii. Duringthe Tokai slow slip event (Ozawa et al., 2002, 2005),Yoshida et al. (2006) reported variations in the slab andcrustal seismicity concurrent with changes in the plate slipvelocity. Liu and Rice (2005) noted that the initiation ofthe large 2001–2002 transient in Guerrero, Mexico,coincided with two normal-faulting earthquakes ofMw=5.0 and 5.8 relatively far inland from the trench,and the transient was followed by several thrust earth-quakes closer to the trench. The 2001–2002 largetransient was suggested to act as a spatial–temporalconnection between the two clusters of seismicity.Repeated episodes of apparent switching betweendown-dip and up-dip earthquake activity have earlierbeen reported by Dmowska et al. (1988) for other regions

that have hosted large earthquakes along the MiddleAmerican Trench (MAT). They suggested earthquakesnear the down-dip end of the locked seismogenic zone as amechanism of communication between these seismicityclusters.

The above observed connections between transientsand seismicity motivated us to systematically investigatetheir relation in a region with long-term continuous geo-detic measurements and abundant seismicity. Guerrero,Mexico, appears to be a well-qualified candidate for thisstudy. Guerrero is along the MAT, where the Cocos platesubducts beneath the North American Plate (Fig. 1). Theconvergence rate varies from53mm/yr to 58mm/yr alongthe trench in directionN33°E (NUVEL1-A,DeMets et al.,1994). No significant seismic energy has been released inthe northwest seismic gap (∼260° to 258.8°E) since therupture of the December 16, 1911 (Ms=7.8) earthquake(Ortiz et al., 2000). The most recent large earthquakes inthe southeast seismic gap (∼261° to 260°E) were in 1907(Ms=7.9) and 1957 (Ms=7.8). These two segments areroughly within the area affected by aseismic transients,and covered by the Guerrero Global Positioning System(GPS) network. The first continuous GPS station inGuerrero was established at CAYA in January 1997.Permanent and campaign instrumentations were installedsubsequently, and now consist of over 10 years of defor-mation record at some sites (Larson et al., 2004).Locations of the permanent stations are shown in Fig. 1.Two large aseismic transients (southward surface dis-placement of∼2 to 5 cm),which lasted for severalmonthsin early 1998 and from October 2001 to April 2002,respectively, have been reported from the continuousmeasurements (Lowry et al., 2001; Kostoglodov et al.,2003). Recently, one transient with a deformation sizecomparable to the 2001–2002 event was detected fromMarch to December 2006 (Larson et al., 2007). CompleteGPS time series of the north-component displacementbetween 1996 and 2004 are shown in Fig. 1 of Lowry(2006).

2. Earthquake catalogs and determination ofcompleteness magnitude Mc

The GCMT (Global Centroid Moment Tensor) andNEIC (National Earthquake Information Center) catalogsare searched for seismic events between 1995 and 2006,within the area of latitude 16° to 20°N and longitude 258°to 262°E, which covers the region affected by the majortransients since 1996, and where the permanent GuerreroGPS network is installed. Earthquake depth is limited tothose less than 100 km, to roughly include events onlyrelated to the shallow subduction process. The relocated

Fig. 1. Tectonic setting of the Guerrero, Mexico, region. Middle American Trench (MAT) defines the plate boundary between the Cocos and NorthAmerican Plates. Bold arrows indicated the direction and magnitude of subduction, based on NUVEL1-A (DeMets et al., 1994). Black diamondsshow the locations of permanent GPS stations. Dashed-line box is along the subduction direction and surrounds the region where most of the Guerreropermanent GPS stations are installed and mainly affected by aseismic transients, and thus the studied area in this paper.

495Y. Liu et al. / Earth and Planetary Science Letters 262 (2007) 493–504

Centennial catalog (Engdahl et al., 1998; Engdahl andVillaseñor, 2002) is also used for better constraints on thelocations of particular earthquakes of interest, as will bediscussed in Section 3.

Among all the events satisfying the above searchcriterions, the lowest magnitude in the GCMT catalog isMw=4.9, and the NEIC catalog includes smaller eventsdown tomb=2.5. However, the number of events at lowermagnitudes may increase with time as an effect of im-proved seismic instrumentation detection capability. Thatinfluences our attempts to identify the natural variations ofseismicity rate. Thus, it is necessary to define a complete-ness (cut-off) magnitudeMc, above which the catalog canbe considered complete and appropriate for this study.There are basically two approaches in the literature todetermine Mc. One is to calculate what the instrumentsshould be able to detect, given their configuration, sensi-tivity, noise level and observations of which earthquakesparticular instruments have or have not detected, follow-ing the procedures in Schorlemmer et al. (2006). This is avery precise but complicated approach for the purpose ofthe present study. As a simple and generally robust

approach, we use the maximum likelihood method (Aki,1965) to calculate the b-value for a wide range of Mc:

b ¼ log10 eð ÞhMi �Mc

¼ 0:4343hMi �Mc

; ð1Þ

where ⟨M⟩ is the mean magnitude of all events equal to orhigher than the detection threshold Mc. For all the NEICearthquakes between 1995 and 2006, with depths less than100 km, within the dashed-line box, b-value, with 98%error bars, versus Mc is shown in Fig. 2(a). b initiallyincreases with Mc when Mc is below the detectabilitythreshold. When Mc is high enough, b becomes statis-tically constant (e.g., b is roughly constant for differentvalues ofMc within the calculation error), and the catalogcan be considered as complete.A completenessmagnitudeof ∼4.3 can be roughly determined. In addition, thefrequency–magnitude distributions, shown in Fig. 2(b),also suggest aMc around the same value. We assume thatthe drop in the number of events within each 0.1-magn-itude bin is caused by the incomplete reporting of events inthe NEIC catalog. Adding a 0.2 or 0.3 safe factor to Mc

Fig. 2. (a) Calculated b-value versus completeness magnitude Mc forall NEIC events, 1995–2006, depth less than 100 km, within thedashed-line box shown in Fig. 1. 98% error bars are shown on b. AsMc increases, the number of analyzed events decreases, and the errorson b becomes larger. b becomes statistically constant for Mc≥4.5.(b) Number of events: cumulative (circle) and within each 0.1-magnitude bin (cross), for the same data set. Decrease of seismicactivity for magnitude lower than ∼3.8 is assumed to be due to theincomplete report of the NEIC catalog. A completeness (cut-off)magnitude of Mc=4.5 is used for subsequent analysis.

496 Y. Liu et al. / Earth and Planetary Science Letters 262 (2007) 493–504

inferred from the statistical analysis [K. Felzer, privatecommunication], we choose a completeness magnitude ofMc=4.5. It is also consistent with an average completenessthreshold of 4.3 to 4.4 reported for the NEIC catalog in amajority of regions [http://earthquake.usgs.gov/regional/neic/neic_bulletins.php]. The Centennial catalog has acut-off magnitude of 5.5 for earthquakes between 1964and April 2002.

In Section 3, we present the seismicity variations be-tween 1995 and 2006, showing the distribution of GCMTand NEIC events, the epicentral distance to trench versustime, and the seismicity rate in each four-year span, i.e.,1995–1998, 1999–2002 and 2003–2006. The earth-quakes for periods including the three large transientsare projected to a vertical cross-section along line AB(Figs. 3, 5 and 7), which is perpendicular to the trench.Weuse the subduction slab geometry, specifically, the thrustinterface profile determined by Kostoglodov et al. (1996)

using the seismicity data of the regional network inGuerrero. The slab profile is also used in the dislocationmodel of Kostoglodov et al. (2003) to fit the observedsurface deformation in the 2001–2002 transient.

3. Seismicity and transients, 1995–2006

3.1. 1995–1998

Fig. 3(a) is themap view of the seismicity fromGCMT(beachballs) and NEIC (gray dots) catalogs between 1995and 1998. Fig. 3(b) shows earthquakes between 1997 and1998 in the dashed-line box projected to a vertical cross-section along AB (red line). For the same earthquake ofinterest, we plot the compressional quadrants of theGCMT beachball, the NEIC and Centennial (if availablein that catalog) locations with the same color.

3.1.1. Possible transient in 1996Before the first permanent GPS station was installed at

CAYA in January 1997, survey measurements were con-ducted in March 1992, September 1995 and April 1996,along the coast and inland of Guerrero. For a completedescription of campaign sites and operation epoches, seeLarson et al. (2004). Evidence of a moderate-size transientwas found from the 1995 and 1996 survey records atACAP (Acapulco). The north component of displacementrelative to the North American Plate is ∼2 cm (Larsonet al., 2004), almost one order of magnitude larger than theaverage horizontal surface deformationduring the northernCascadia aseismic transients (Dragert et al., 2001).However, we need to be cautious on the identification ofthis transient based on the campaign data at the singlestation, as transient motions can also result from theinstability of the monument, localized mass-wastingphenomena, or different conditions in campaign epoches.The possibility that the slow slip was triggered by theMw=7.3, September 1995Copala earthquake (“091405C”in Fig. 3(a)),∼150 km east of ACAP, cannot be ruled out.

Fig. 4(a) shows the temporal variation of the epicentraldistance to the trench of NEIC events greater than 4.2 inthe dashed-line box between 1995 and 1998. Dashed-linecircles represent earthquakes with normal- (marked byletter “N”) and thrust-faulting (marked by letter “T”)GCMT focal mechanisms. The duration of the 1996transient is poorly constrained due to limited measure-ments. We mark it roughly from November 1995 toFebruary 1996, in Fig. 4, based on the estimate by Larsonet al. (2004). Since the beginning of 1995, no earthquakeswith magnitude higher than 4.5 were reported in eithercatalog until December 20, 1995, when a Mw=5.3normal-faulting earthquake occurred near the border

Fig. 3. (a) Map view of seismicity in the Guerrero region, 1995–1998. Beachballs show the centroid locations and focal mechanisms of GCMTevents. Labels on top are in order “Month/Day/Year/Event of that day”. Gray dots show the epicenters of NEIC events larger than the completenessmagnitudeMc=4.5. Dot size is proportional to event magnitude. GCMT and NEIC epicenters sometimes deviate by tens of km. NEIC events withinthe dashed-line box are used in the seismicity analysis. For reference, black diamonds represent locations of permanent GPS stations. (b) Seismicity,1997–1998, within the dashed-line box projected to a vertical cross-section along AB (red line). Subduction thrust interface is adopted fromKostoglodov et al. (1996, 2003) and Manea et al. (2004). The blue dot and beachball with blue compressional quadrants are NEIC and GCMTlocations of the extensional earthquake “042098B”, respectively. Blue star represents its position from the relocated Centennial catalog (Engdahlet al., 1998; Engdahl and Villaseñor, 2002).

497Y. Liu et al. / Earth and Planetary Science Letters 262 (2007) 493–504

between Guerrero and Michoacán, with an epicentraldistance of∼170 km inland from the trench and a depth of78 km (NEIC). A similar depth of 76 km is determined inGCMT. This extensional earthquake corresponded to theinitiation of the possible 1996 transient, and occurred inthe subducting slab. A thrust earthquake of Mw=5.5

occurred on April 23, 1996, shortly after the transient,given the estimated duration. While the GCMT centroidlocation of the April event lies on the trench, NEIC reportsan epicenter ∼40 km inland from the trench, which isprobably a more precise estimate using data from theregional seismic network in Guerrero. Another cluster of

Fig. 4. (a) Spatial–temporal variation of NEIC events with magnitude greater than 4.2, within the dashed-line box, 1995–1998. Circle size isproportional to event magnitude. Dashed-line circles are NEIC events that have GCMT solutions: normal-faulting marked by letter “N”; thrust-faulting marked by letter “T”. Numbers in the parenthesis are moment magnitudes from GCMT. Only an average number is marked for a cluster ofearthquakes, e.g., “T (∼5.3)” for the five thrust-faulting earthquakes after the 1998 transient. Gray circles represent events belowMc=4.5, but greaterthan 4.2. (b) Number of earthquakes in every 10 days, 1995–1998. Black bars show numbers of events greater thanMc=4.5. Gray bars show numbersof events between 4.2 and 4.5. Two light gray strips approximately mark the durations of seismic transients in, possibly, 1996, and 1998.

498 Y. Liu et al. / Earth and Planetary Science Letters 262 (2007) 493–504

thrust events was reported in middle July, with the largestmagnitude of Mw=6.6.

3.1.2. Transient in 1998About 1 year after the first permanent GPS station was

installed at CAYA, a transient motion, starting from early1998, was observed and lasted for ∼5 months. Thereversed motion was later confirmed by displacement atan inland station POSW. Continuous measurements fromthe beginning of 1997 to late 2000 were used tomodel theaseismic deformation and suggest a total static displace-ment of 2 mm east, 26 mm south and 16 mm up duringthis transient (Lowry et al., 2001). Along-strike propaga-tion, a feature like that exhibited by the Cascadiatransients (Dragert et al., 2001), is also implied based onthe surface eastward deflection at the beginning andwestward deflection at the end of the transient; a simplestatic slip patch cannot duplicate such a time-varyingfeature. The deflection signal is consistent with a NW toSE slip propagation motion, which is also suggested bythe seismicity variation associated with this transient.

During the transient slip period, an extensional earth-quake of Mw=5.9 occurred on April 20, 1998, on theNW border between Guerrero and Michoacán, near theepicenter of the 1995 normal-faulting earthquake.GCMT, NEIC and Centennial locations of this Aprilevent are shown in Fig. 3 by a beachball with blue

compressional quadrants, blue dot and blue star,respectively. There is a significant discrepancy of morethan 40 km between the GCMT and NEIC horizontallocations, and NEIC has a better agreement with therelocated Centennial position. Earthquakes with magni-tudes less than 4.5, shown as gray circles in Fig. 4(a),near the extensional earthquake epicentral area were alsoreported preceding or at the early stage of the 1998transient. Furthermore, a group of thrust-faulting earth-quakes occurred near Acapulco in July 1998, immedi-ately after the transient. It is clear from Figs. 3 and 4 thatthe 1998 transient coincide with an extensional earth-quake in the subducting slab at NW Guerrero and wasfollowed by a cluster of shallower thrust earthquakesclose to the trench near Acapulco, suggesting a NW toSE slow slip propagation and up-dip stress transfer.

Fig. 4(b) shows the number of seismic events, abovethe completeness magnitude 4.5 (black bars) andbetween 4.2 and 4.5 (gray bars), per 10 days for theexamined 4-year period. High seismic rate is generallyobserved in the temporal vicinity of the transient slipevents.

3.2. 1999–2002

Fig. 5(a) is themap view of the seismicity fromGCMTand NEIC catalogs between 1999 and 2002. Fig. 5(b)

Fig. 5. (a) Map view of seismicity in Guerrero, 1999–2002. Extensional earthquakes “100801B” and “102901B” are colored in blue and green,respectively. Thrust event “041802B” is colored in orange. NEIC and Centennial locations of “100801B”, represented by blue dot and star, almostoverlap. (b) Seismicity, 2001–2002, within the dashed-line box projected to a vertical cross-section along AB. Symbol representations are the same asin Fig. 3.

499Y. Liu et al. / Earth and Planetary Science Letters 262 (2007) 493–504

shows the earthquakes between 2001 and 2002 within thedashed-line box projected to the vertical cross-sectionalong AB.

3.2.1. Large transient in 2001–2002From early 1999 to late 2001, GPS measurements

vaguely suggest three small aseismic transients with the

north-component of displacement less than 2 mm, atleast an order of magnitude smaller than that of the 1998transient (Lowry, 2006). Seismicity during the sameperiod is relatively sparse, as shown in Fig. 6.

The Guerrero region became seismically more activesince late 2001, signifying the beginning of a largetransient. Aseismic deformation is clearly visible on the

500 Y. Liu et al. / Earth and Planetary Science Letters 262 (2007) 493–504

time series of all permanent GPS stations then operating(Kostoglodov et al., 2003). The reversed motion was firstdetected at stations ACAP, CAYA, IGUA and YAIG, nearthe border of the NW and SE seismic gaps, then about2 months later, at a northwest station ZIHP and southeaststations PINO, OAXA (∼120 km northeast of PINO, notshown in Fig. 1). The temporal delay in the transientmotion onsets at different stations suggests a bilateralpropagation at a speed of about 6–9 km/day (Kostoglodovet al., 2003), similar to the speed inferred for the northernCascadia and southwest Japan short-term slow slipevents. Anomalous surface deformation was observedfrom October 2001 to April 2002 over an area of morethan ∼550×250 km2, resulting in an equivalent momentmagnitude of ∼7.5.

A cluster of earthquakes relatively far inland from thetrench coincided with the beginning of the transient. Twoof them (“100801B” and “102901B” in Fig. 5) arenormal-faulting events with Mw=5.8 and 5.0. While theGCMTand NEIC catalogs are ambiguous in the depths ofthe two extensional earthquakes, a study reported in anabstract by Pacheco et al. (2002) suggests the October 8,2001 Coyuca earthquake occurred at a shallow depth of8 km and is thus a crustal event. GCMT and NEIC reportdepths at 10 and 15 km, respectively, and are suspected tobe fixed depths in both catalogs. As shown in Fig. 5(b),the GCMT location of “100801B” is shifted by 30–40 km

Fig. 6. (a) Epicentral distance from trench and (b) seismicity rate,1999–2002. Symbol representations are the same as in Fig. 4. NEICevents above the completeness magnitudeMc=4.5 are shown here. “T(max 6.7)” represents the largest moment magnitude among the thrust-faulting earthquakes following the 2001–2002 transient is 6.7. Thetransient marked from late 2002 continues on Fig. 8.

inland from its NEIC epicenter (blue dot). The Centennialcatalog (blue star) and records from the GuerreroAccelerograph Network stations [http://www.seismo.unr.edu/Guerrero/] also suggest epicenters close to theNEIC location. The normal-faulting mechanism might beexplained by the shallow extensional stresses left in thewake of indentation of the upper plate by the locallysteepened section of the slab near 80 km from the trench.The possibility that the event “100801B”might have beentriggered by the transient is also mentioned by Kostoglo-dov et al. (2003), although their discussion would seem torequire an offshore nucleation of the transient slip.Nevertheless, the epicenters of the two October earth-quakes are approximately along the same trench-normalline inland from stations CAYA, ACAP, where thetransient episode started. This spatial–temporal correla-tion provides evidence that stressing from the nearbyseismicity may have triggered the transient or had acommon origin with it. The seismic rates became higherduring the transient; the October 8 Coyuca earthquakeproduced a large number of aftershocks (N300) that lasted∼6 months, overlapping the duration of the transient(Kostoglodov et al., 2003). We also note that toward theend of the transient, in middle April 2002, several thrust-faulting earthquakes occurred close to the trench, morethan 100 km west of stations CAYA, ACAP. This is alsoconsistent with the bilateral propagation of the slow slipevent. The largest magnitude of the thrust events isMw=6.7 (“041802B” in Fig. 5). Although, GCMT andNEIC horizontal locations have a∼30 km discrepancy inthe trench-normal distance, the along-strike locations arerelatively well resolved, thus would not affect theconsistency with the slow slip propagation direction.

3.3. 2003–2006

Fig. 7(a) is themap view of the seismicity fromGCMTand NEIC catalogs between 2003 and 2006. Fig. 7(b)shows the earthquakes between 2005 and 2006 within thedashed-line box projected to the vertical cross-sectionalong AB.

3.3.1. Two possible small transients in 2003 and 2004Lowry (2006) also inferred a small aseismic transient

from late 2002 to early 2003, as marked on Fig. 6 andcontinued on Fig. 8. The deformation signal is mostprominent at coastal stations ZIHP and CAYA, which arein the epicentral area of many small earthquakes duringthat period, suggesting a casual relation between theseismicity and transient. We cannot rule out the possibilitythat it is an aftermath of the large transient in 2001–2002,as most of the seismicity in late 2002 and early 2003 are

Fig. 7. (a) Map view of seismicity in Guerrero, 2003–2006. Extensional earthquakes “121405A”, “022006A” and “081106A” are colored in blue,green and orange, respectively. A recent thrust earthquake “041307A” following the 2006 transient is also shown. (b) Seismicity, 2005–2006, withinthe dashed-line box projected to a vertical cross-section along AB. Legends are the same as in Fig. 3.

501Y. Liu et al. / Earth and Planetary Science Letters 262 (2007) 493–504

close to the trench and many have thrust-faulting focalmechanisms.

An even smaller transient was inferred to have oc-curred in early 2004 (Lowry, 2006), as marked in Fig. 8.We do not discuss that in detail, due to the little constrainton deformation. Similarly, seismicity rate was high duringand after this possible transient, with a majority of earth-quakes close to the trench.

3.3.2. Large transient in 2006The most recently detected large transient in the

Guerrero region started aroundMarch 2006, and the GPSsignal began to return to the normal trend at some stationsin late September while extending into December atothers (Larson et al., 2007). The size of this transient iscomparable to that in 2001–2002; the total horizontaldisplacement at CAYA is about 6 cm. The reversed

Fig. 8. (a) Epicentral distance from trench and (b) seismicity rate,2003–2006. NEIC events above 4.2 are shown here. Symbolrepresentations are the same as in Fig. 4. The transient marked inearly 2003 is continued from late 2002, as shown in Fig. 6.

502 Y. Liu et al. / Earth and Planetary Science Letters 262 (2007) 493–504

motion was continuously observed at all permanent GPSstations, except ZIHP where only a net displacementbefore and after the transient is obtained due to technicalproblems. Larson et al. (2007) modeled the deformationwith four patches of rectangular fault planes, and foundthe east-component of the slip anomaly can be dividedinto two stages. In the first stage, from February to June2006, stations (CAYA,COYU,ACAP,ACYAandCPDP)near the coast experienced a faster eastward movement.That was followed, in the second stage, by a westwardmotion from May to the end of this transient episode. Onthe contrary, inland stations (MEZC, IGUA and YAIG)moved faster toward the west in the first stage andcontinued to move westward but with gradually decreas-ing slip rates.

The 2006 large transient was preceded by a cluster ofearthquakes far inland on the NW border of Guerrero, inthe same epicentral area of the extensional earthquakes in1996 and 1998. The earliest among the cluster of GCMTevents was a Mw=4.9, normal-faulting earthquake onDecember 14, 2005 (“121405A” in Fig. 7, colored blue),about 2months before the transient. On February 20, 2006,shortly before the transient signal could be detected byGPS, a Mw=5.2 event (“022006A”, colored green)occurred at roughly 150 km (NEIC) inland from thetrench. One month later, another normal-faulting eventwith a similar magnitude was reported in GCMT with itscentroid location slightly NW out of the dashed-line box(“032006A”). More earthquakes with large distances to the

trenchwere observed during the transient; the largestwith aMw=6.0 onAugust 11, 2006(“081106A”, colored orange).All of the three normal-faulting earthquakes in February,March and August 2006 are located at depths ∼60 km ordeeper thus in the subducting slab. The GCMT and NEIChorizontal locations agree relatively well with each other,except for event “022006A”. Toward the end of thetransient, the seismicity cluster seemed to migrate closer tothe trench. Although all events are too small to haveGCMT solutions, their locations suggest a shallow thrustfocal mechanism. We also plot in Fig. 7 a recentMw=5.9thrust earthquake on April 13, 2007 (“041307A”), whichlies southeast from the extensional earthquakes clusterbefore and during the transient, and is much closer to thetrench. Although the GCMT solution suggests the thrustfault plane activated during event “041307A” inclineseither toward the ocean or toward the continent at a steeperangle than the subducting slab, the spatial evolution ofseismicity cluster is consistent with a northwest towardsoutheast aseismic slip migration implied from the tran-sient duration offsets at different GPS stations (Larsonet al., 2007).

4. Conclusion and discussion

Recent observations of aseismic deformation transi-ents and sometimes associated deep non-volcanic tremorsin the circum-Pacific subduction zones pose significantquestions as to their origin, and also relative to existingconcepts of interseismic loading of the locked seismo-genic regions. Stressing in the up-dip seismogenic zone isincreased episodically due to down-dip transient slips, andit can be made more prone to failure in a large thrustearthquake. Thus, it is important for seismic hazardassessment to search and identify patterns of spatiotem-poral seismicity variation associated with transients. Werejuvenate the suggestions made by Dmowska et al.(1988), which could not be linked to a convincingmechanism at that time, on possible communicationbetween extensional seismicity clusters down-dip in theslab, and later thrust clusters in the shallow seismogeniczone, along theMiddle American Trench offMexico. Thepattern seems to continue in the recent seismicity alongother region ofMAT (Guerrero) thatwe have studied here.

We searched the GCMT and NEIC catalogs forearthquakes in a twelve-year period (1995–2006) in thearea affected by the aseismic transients in Guerrero,Mexico. The seismicity variation patterns are identified tobe spatial–temporally associated with the transientsobserved by the Guerrero GPS network since 1996.Three large transients in 1998, 2001–2002 and 2006 areall correlatedwith high seismic rates in the studied area. In

503Y. Liu et al. / Earth and Planetary Science Letters 262 (2007) 493–504

particular, we found that the initiation of the transientsoccurs in association with a cluster of extensional earth-quakes relatively far inland from the trench, in thesubducting slab or the overlying crust, and may befollowed by a cluster of shallow earthquakes close to thetrench, among which many have thrust-faulting mech-anisms. In some cases, such as the transient in 2001–2002, both types of activity are found bracketing thetransient period. The beginning of the 2006 transientcoincided with two normal-faulting earthquakes inFebruary and March, 2006, near the northwest border ofGuerrero. Toward the end of the transient, the NEICcatalog shows the seismicity cluster moved closer to thetrench, implying hypocenters up-dip in the seismogeniczone.

The assembled evidence suggests that aseismic defor-mation transients may serve as a mechanism of stresscommunication between distant regions, e.g., down-dipand up-dip, in subduction zones. The Guerrero transientsseem to be initiated by earthquakes far inland from thetrench, in the subducting slab or the continental crust, or tohave a common cause of that activity. They transferstresses to the locked shallow part in a manner whichsometimes results in thrust earthquakes there. Thatconjecture has been taken into account in the numericalmodeling of subduction earthquakes and aseismictransients using the rate and state-dependent friction(Liu and Rice, in press).When amoderate, step-like stressperturbation, e.g., from a nearby earthquake, is applied tothe thrust interface, sequential aseismic transients can beresulted, and the timing of the next large thrust earthquakeis affected by three factors, namely, when, where and howlarge is the stress perturbation.

The discovery of aseismic deformation transients isan important development in our knowledge of theseismic cycle along major plate boundaries. It posessignificant puzzles and changes the way we should thinkabout the loading of seismogenic zones. Such transientscontribute episodic steps in loading to the thrustinterface. Their improved understanding seems likely,based on observations for the MAT and on theory, toincrease the predictability of earthquakes.

Acknowledgments

This work was supported at Harvard by NSF-EARaward 0510196 and USGS grants 06HQGR0047 and07HQGR0057, and at the University of Colorado byNSF-EAR award 0609646. We thank Karen Felzer fordiscussion on the catalog completeness magnitude,Renata Dmowska and Meredith Nettles for commentson various aspects of the study, Vladimir Kostoglodov

and Jose Antonio Santiago (UNAM) for support of theGPS network in Guerrero, Aron Meltzner and KerrySieh for discussion on the interpretation of Sumatracoral microatoll records, and Göran Ekström andRoland Bürgmann for comments which helped toimprove the original presentation. The Generic MappingTool (GMT) is used to plot some figures in this paper.

References

Aki, K., 1965. Maximum likelihood estimate of b in the formula logN=a−bM and its confidence limits. Bull. Earthq. Res. Inst. 43,237–239.

DeMets, C., Gordon, R.G., Argus, D.F., Stein, S., 1994. Effect ofrecent revisions to the geomagnetic reversal time on estimates ofcurrent place motions. Geophys. Res. Lett. 21 (20), 2191–2194.

Dragert, H.,Wang,K., James, T.S., 2001.A silent slip event on the deeperCascadia subduction interface. Science 292 (5521), 1525–1528.

Dmowska, R., Rice, R.J., Lovison, L.C., Josell, D., 1988. Stress transferand seismic phenomena in coupled subduction zones during theearthquake cycle. J. Geophys. Res. 93 (B7), 7869–7884.

Douglas, A., Beavan, J., Wallace, L., Townend, J., 2005. Slow slip onthe northern Hikurangi subduction interface, New Zealand.Geophys. Res. Lett. 32 (L16305). doi:10.1029/2005GL023607.

Engdahl, E.R., Villaseñor, A., 2002. Global seismicity: 1900–1999.In: Lee, W.H.K., Kanamori, H., Jennings, P.C., Kisslinger, C.(Eds.), International Handbook of Earthquake and EngineeringSeismology, Part A. Academic Press, pp. 665–690. Chapter 41.

Engdahl, E.R., vander Hilst, R., Buland, R., 1998. Global teleseismicearthquake relocation with improved travel times and proceduresfor depth determination. Bull. Seismol. Soc. Am. 88, 722–743.

Hirose, H., Obara, K., 2005. Repeating short- and long-term slow slipevents with deep tremor activity around the Bungo channel region,southwest Japan. Earth Planets Space 57 (10), 961–972.

Hirose, H., Hirahara, K.,Kimata, F., Fujii, N.,Miyazaki, S., 1999. A slowthrust slip event following the two 1996 Hyuganada earthquakesbeneath theBungoChannel, southwest Japan. Geophys. Res. Lett. 26(21), 3237–3240.

Kao, H., Shan, S.-J., Dragert, H., Rogers, G., Cassidy, J.F., Ramachan-dran, K., 2005.Awide depth distribution of seismic tremors along thenorthern Cascadia margin. Nature 436 (7052), 841–844.

Kostoglodov, V., Bandy, W., Dominguez, J., Mena, M., 1996. Gravityand seismicity over the Guerrero seismic gap. Geophys. Res. Lett.23 (23), 3385–3388.

Kostoglodov, V., Singh, S.K., Santiago, J.A., Larson,K.M., Lowry, A.R.,Bilham, R., 2003. A large silent earthquake in the Guerrero seismicgap, Mexico. Geophys. Res. Lett. 30 (15), 1807. doi:10.1029/2003GL017219.

Larson, K.M., Lowry, A.R., Kostoglodov, V., Hutton, W., Sánchez, O.,Hudnut, K., Suárez, G., 2004. Crustal deformation measurementsin Guerrero, Mexico. J. Geophys. Res. 109 (B04409). doi:10.1029/2003JB002843.

Larson, K.M., Kostoglodov, V., Miyazaki, S., Santiago, J.A., 2007.The 2006 aseismic slow slip event in Guerrero, Mexico: Newresults from GPS. Geophys. Res. Lett. 34 (L13309). doi:10.1029/2007GL029912.

Liu, Y., Rice, J.R., 2005. Aseismic slip transients emerge spontaneouslyin three-dimensional rate and state modeling of subduction earth-quake sequences. J. Geophys. Res. 110 (B08307). doi:10.1029/2004JB003424.

504 Y. Liu et al. / Earth and Planetary Science Letters 262 (2007) 493–504

Liu, Y., Rice, J.R., in press. Spontaneous and triggered deformationtransients in a subduction fault model. J. Geophys. Res. doi:10.1029/2007JB004930.

Lowry, A.R., 2006. Resonant slow fault slip in subduction zonesforced by climatic load stress. Nature 442 (7104), 802–805.

Lowry, A.R., Larson, K.M., Kostoglodov, V., Bilham, R., 2001.Transient fault slip in Guerrero, southern Mexico. Geophys. Res.Lett. 28 (19), 3753–3756.

Manea, V.C., Manea, M., Kostoglodov, V., Currie, C.A., Sewell, G.,2004. Thermal structure, coupling and metamorphism in theMexican subduction zone beneath Guerrero. Geophys. J. Int. 158(2), 775–784.

Mazzotti, S., Adams, J., 2004. Variability of near-term probability forthe next great earthquake on the Cascadia subduction zone. Bull.Seismol. Soc. Am. 94 (5), 1954–1959.

Meltzner, A.J., Sieh, K., Briggs, R., Chiang, H.-W., Shen, C.-C.,Suwargadi, B.W., 2007. Coseismic, postseismic, and interseismicdeformation at the boundary of the 2004 and 2005 mega thrustruptures: insights from coral microtolls. Seismol. Res. Lett. 78 (2),300.

McCausland, W., Malone, S., Johnson, D., 2005. Temporal and spatialoccurrence of deep non-volcanic tremor: from Washington tonorthern California. Geophys. Res. Lett. 32 (L24311). doi:10.1029/2005GL024349.

Murray, J.R., Segall, P., 2005. Spatiotemporal evolution of a transientslip event on the San Andreas fault near Parkfield, California.J. Geophys. Res. 110 (B09407). doi:10.1029/2005JB003651.

Natawidjaja, D.H., Sieh, K., Ward, S.N., Cheng, H., Edwards, R.L.,Galetzka, J., Suwargadi, B.W., 2004. Paleogeodetic records of seismicand aseismic subduction from central Sumatran microatolls, Indone-sia. J. Geophys. Res. 109 (B04306). doi:10.1029/2003JB002398.

Natawidjaja, D., Sieh, K., Galetzka, J., Suwargadi, B., Cheng, H.,Edwards, R.L., 2007. Interseismic deformation above the Sundamegathrust recorded in coral microatolls of the Mentawai islands,West Sumatra. J. Geophys. Res. 112 (B02404). doi:10.1029/2006JB004450.

Ohta, Y., Freymueller, J.T., Hreinsdottir, S., Suito, H., 2006. A largeslow slip event and the depth of the seismogenic zone in the southcentral Alaska subduction zone. Earth Planet. Sci. Lett. 247 (1-2),108–116.

Ortiz, M., Singh, K.S., Kostoglodov, V., Pacheco, J., 2000. Sourceareas of the Acapulco-San Marcos, Mexico earthquakes of 1962(Mw=7.1, 7.0) and 1957 (Mw=7.7), as constrained by tsunami anduplift records. Geofís. Int. 39, 337–348.

Ozawa, S., Murakami, M., Kaidzu, M., Tada, T., Sagiya, T., Hatanaka,Y., Yarai, H., Nishimura, T., 2002. Detection and monitoring ofongoing aseismic slip in the Tokai region, central Japan. Science298 (5595), 1009–1012.

Ozawa, S., Murakami, M., Kaidzu, M., Hatanaka, Y., 2005. Transientcrustal deformation in Tokai region, central Japan, until May 2004.Earth Planets Space 57 (10), 909–915.

Pacheco, J.F., Iglesias, A., Singh, S.K., 2002. The 8 October Coyuca,Guerrero, Mexico earthquake (Mw 5.9): A normal fault in theexpected compressional environment. Seismol. Res. Lett. 73 (2),263.

Rogers, G., Dragert, H., 2003. Episodic tremor and slip on theCascadia subduction zone: the chatter of silent slip. Science 300(5627), 1942–1943.

Segall, P., Desmarais, E.K., Shelly, D., Miklius, A., Cervelli, P., 2006.Earthquakes triggered by silent slip events on Kilauea volcano,Hawaii. Nature 442 (7098), 71–74.

Schorlemmer, D., Woessner, J., Bachmann, C., 2006. ProbabilisticEstimates of Monitoring Completeness of Seismic Networks,100th Anniversary Earthquake Conference.

Shelly, D.R., Beroza, G.C., Ide, S., Nakamula, S., 2006. Low-frequency earthquakes in Shikoku, Japan, and their relationship toepisodic tremors and slip. Nature 442 (7099), 188–191.

Szeliga, W., Melbourne, T.I., Miller, M.M., Santillan, V.M., 2004.Southern Cascadia episodic slow earthquakes. Geophys. Res. Lett.31 (L16602). doi:10.1029/2004GL020824.

Wallace, L.M., Beavan, J., 2006. A large slow slip event on the centralHikurangi subduction interface beneath the Manawatu region,North Island, New Zealand. Geophys. Res. Lett. 33 (L11301).doi:10.1029/2006GL026009.

Yoshida, A., Hososno, K., Tsukakoshi, T., Kobayashi, A., Takayama,H., Wiemer, S., 2006. Change in seismic activity in the Tokairegion related to weakening and strengthening of the interpolatecoupling. Tectonophysics 417, 17–31.